The rise of the AI ​​data center does not go bankrupt, but the ‘pause’ is in a trend

Expanding and sedimentation data centers have become a basic element of suburbs in some places such as shopping centers and soccer fields. However, when Microsoft pulled the plug In the planned data centers in Ohio last month, the questions about whether the dough of the data center had already broken. A Wells Fargo report last Monday by saying that some data centers planned by Amazon Web Services were being reconsidered to market anxiety.

But the bust may have finished before it began. And in any case, a “pause” in some data centers projects is within an expense environment that remains strong.

“We continue to see the accelerated scale of AI implementations throughout the data centers market, with strong demand signals that reinforce our growth in the short and long term level,” said Giordano Albertazzi, CEO of Ohio Data Center to provide Verb in a earning call last week. His shared ended the week with 22%.

Amazon and Nvidia reaffirmed last week that the data centers market is still strong.

“There has really been no significant change,” said Kevin Miller, vice president of Amazon Global Data Centers, at a conference organized by the Hamm Institute for American Energy. “We continue to see a very strong demand, and we are looking for both in the next two years and in the long term and seeing that the numbers only rise.”

That does not mean that strategic thinking about how, where and when to spend exactly does not change as the market evolves and advances should be digested. In the period of six weeks this year, the Deepseek of China broke into the scene, the Stargate initiative of $ 500 billion of $ 500 billion AI was announced, and concerns about rates and commercial wars were announced in markets.

“All that has created a scenario in which the data centers industry is taking a pause, in general,” said Pat Lynch, executive managing director of the data center solutions of the CBRE Commercial Real Estate Company. “I think it is a temporary pause,” added Lynch, noting that the project pipe and its funnel remain significant and CBRE continues to execute agreements. “I’m still cautiously optimistic about future demand, particularly when you think of great AI training models,” Lynch said.

Microsoft had promised an investment of $ 1 billion in OHIO headquarters in the same area where Intel has planned chips factories, but the timeline has slowed.

“After a careful consideration, we will not advance with our plans to build data centers at Licking County sites at this time. We will continue to evaluate these in -line sites with our investment strategy,” said a Microsoft spokesman in a statement to CNBC.

An UBS report last week concluded that among all possible explanations for the cancellations of the data centers, it was very likely that Microsoft had committed in the middle of the AI ​​Rush, and now it was concentrating on the projects that currently make more sense. He pointed out that the Microsoft leased capex rose 6.7 times within two years, with lease obligations of approximately $ 175 billion. “Microsoft bought as much leased data center available as it could in 2022-2024 and now has the visibility of eliminating some of these ‘projects in the initial stage,” UBS wrote. “We found the least support for the explanation of ‘Lull of demand’,” added his report.

Anat Ashkenazi, Alphabet CFO described the cloud supply demand environment as “tight” after its last profits on Thursday. “We could see the variability in the growth rates of income in the cloud depending on the implementation of the capacity each quarter,” he said. “We expect a relatively greater deployment of capacity around 2025.”

“We are not seeing a withdrawal of demand but a strategic reallocation,” said John Carrafiell, Co-Coo de BGO, a world manager of real estate investments with $ 83 billion in assets under administration, including an important portfolio of data centers. The most important players, he says, are not going back, with Microsoft, Google, Meta and Amazon plan to spend more than $ 300 billion in Capex this year, in general, linked to the infrastructure of AI. And, he says, it does not include other important players, such as Operai and Oracle, both involved in the Stargate project.

“Instead of a bust, this is a reorganization of the roof in an environment where power in particular, together with fiber, water and earth, are scarce and strategic,” said Carrafielll. The long -term business adoption will boost the demand for the demand for AI and the data center for the next decade. “We are not even in the first post,” he said.

Power is the soul of data centers, but data centers are not operations plug and play, which require large amounts of electricity for computer energy and fans to maintain fresh infrastructure. As the generative adoption of AI passes from early experimentation to the business scale application, the need for low -latency efficiency data centers near the end users will intensify, but it will take time for the correct set of conditions to align with the expected square feet of the data center.

“The new data centers increase in size so dramatically that the network cannot be kept up to date,” said Allan Schur, commercial director of the Micorred Enchathed Rock developer. Three years ago, a large data center was 60 megawatts, sufficient power to supply 20,000 houses, but now it says that the new data centers support all the uses of artificial intelligence are requesting 500 megawatts or more.

This rapid growth in the use of electricity is in addition to the new demand for the manufacture and electrification of transport, which together weigh on supply and infrastructure. Data centers pose a unique challenge for public services, which must guarantee that they can provide energy to all customers, even in times of demand peak. “That is why some profits cite long interconnection waiting times for data centers,” said Schur. “Public services must invest in new substations and may also need to expand transmission and generation, all of which has been,” he added.

CBRE has seen the data centers that comprise 2% of their portfolio in 2022 to 10% in 2024, and Lynch hopes that it will continue to grow, and the proximity of power is promoting the current market, since the data centers builders look for areas with access to an abundant power. Georgia, Texas and Ohio mark many of the boxes that builders are looking for, and if an area does not have the capacity of the grid or infrastructure, it must be able to climb quickly.

“Having great energy availability within 36 months is attractive to customers,” Lynch said.

According to Datacenters.com, three percent of the world’s power is linked to data centers.

Schur said that Uneced Rock data indicates that there is a lot of energy available to meet demand, most of the time. Of the 8,760 hours a year, the grid is only under stress due to a fraction of them. “If we can relieve the demand of the network for those 100 to 500 hours, long interconnection delays can be shortened,” he said.

An important distinction can be made between the idea of ​​a broader deceleration and some of the recent pauses promulgated by the main technology companies, according to McKinsey & Company’s main partner, Pankaj Sachdeva, which investigates the development of the data center and expects a flow and flow.

Based on the recent McKinsey modeling, which does not include the impact of rates, the data center market is expected to grow in the range of 20% to 25% in the next five to seven years, but year after year there will be variations in the growth rate. “It won’t be linear,” he said.

Changes in the rate will introduce new cost pressures in the AI ​​supply chains and data centers, particularly with critical mineral tariffs on the horizon.

“These interruptions will raise hardware costs, the impact supply strategies and require companies to rethink their long -term acquisition models,” said John Archer, director of Senior Delivery and leader of transformation of the supply chain in Slalom Consulting. In the short term, IA and Cloud suppliers will need to implement cost mitigation strategies, such as renegotiating supplier contracts and inventory optimization.

“It can be expected that a long -term impulse towards geographical diversification, joint manufacturing in friendly regions with rates and the deepest integration of the supply chain analysis promoted by AI adapt to evolving commercial policies,” said Archer.

A factor that has not changed is that the calculation power is currently expensive, and much more is needed for the AI ​​software and hardware, according to Suresh Venkatesan, CEO of Poet Technologies, a company that quotes in the stock market that develops power solutions for data centers. “The explosion in AI challenges data centers to find more efficient solutions because AI requires calculation power in such a volume that is different from everything we have witnessed,” he said. “While a data center project can hit a wall, it is likely that others arise, because there are no indications of a deceleration in connectivity demand,” he added.